An intelligent evidence-theory-based structural reliability analysis method based on convolutional neural network model
LIU Xin,Jun Wan,Weiqiang Jia,Xiang Peng,WU Shaowei,LIU Kai,Xin LIU,Shaowei WU,Kai LIU
DOI: https://doi.org/10.1016/j.cma.2024.116804
IF: 6.588
2024-03-01
Computer Methods in Applied Mechanics and Engineering
Abstract:Combined with the convolutional neural network (CNN) model, an intelligent structural reliability analysis method based on evidence theory is developed to improve the accuracy and efficiency of reliability analysis. Firstly, the uncertainties in engineering structures are described by the Frame of Discernment (FD) and the Basic Probability Assignment (BPA). Secondly, the Optimal Latin Hypercube Design (OLHD) is adopted to obtain the sample focal elements and the Sequential Quadratic Programming (SQP) method is utilized to carry out the extremum analysis of the sample focal elements. Thirdly, the sample focal elements can be reclassified as the belief focal element, the intersect focal element and the failure focal element. Fourthly, the local-densifying method of sample space is applied to ensure the uniformity of sample focal elements and the convolution neural network model is trained by these focal elements. Then, the classifications of the undetermined focal elements could be obtained by the convolution neural network model and the confidence interval can be obtained based on the classification results of all focal elements. Finally, two numerical examples and one engineering application are introduced to investigate the effectiveness of the proposed method.
mechanics,engineering, multidisciplinary,mathematics, interdisciplinary applications